DART-based Data Assimilation for Wind Prediction
In this project sponsored by Siemens, in collaboration with the Lawrence Livermore National Laboratory (LLNL), NCAR/RAL has developed a wind prediction system for wind energy applications based on state-of-the-science EnKF algorithms, available in the Data Assimilation Research Testbed (DART) community facility (Anderson et al. 2009). Several different kinds of datasets can be ingested into DART including marine data (WMO/GTS); satellite-derived stratospheric winds; radar data; wind profilers; surface meteorological stations (MESONETs); aircraft data (ACARS); sea surface temperature data from satellite, buoys and ships; and meteorological data collected at wind farms.
The simulation domain is centered over Nysted, Denmark. Figure 1 shows the chosen configuration of the domains, where each ensemble member is run over two nested domains with 27 and 9 km horizontal grid increments. In this example, DART was run with 30 ensemble members and by cycling the data assimilation procedure every 6 hours. The test period was 24 hours in duration, starting 01 May 2007.
Figures 2 and 3 show the domain-average vertical profiles of the root-mean-square-error (RMSE) and bias for the north-south component of wind, respectively. When the background information (prior) provided by the 6-hour ensemble forecast is updated with the observations (posterior) both RMSE and bias are considerably reduced, resulting in analysis much closer to the observations than the background forecast.